Timeseries Models 11-22-2024

School Appropriation 1998-2025 with Timeseries Forecast through 2039.

This is a timeseries version of a forecast model presented at the last meeting.
Timeseries models are standard tools in financial forecasting.
They have been around for many years in the form of ARIMA (Autoregressive Integrated Moving Average) or classical time series models.
A Gaussian Process model is often the timeseries of choice for 21st century forecasting.
With timeseries, the number of parameters does not increase with the number of years ahead to forecast. This is a significant advantage over projection methods and makes longer term forecasts possible.
Another advantage of timeseries models is that you can validate them by turning the clock back and forecasting years you already know the result for.

The standard reference for Gaussian Process models.

Gaussian Process models are in many ways an improvement over classical ARIMA models.
They require much less data and make fewer assumptions than classical timeseries, and far fewer assumptions than projections.
Unlike classical time series and linear models, a Gaussian Process model does not need to assume a specific function for the data model.
Usually the only assumption on the form of the function is either that is continuous or smoothly continuous.
The 2006 Rasmussen text is the standard reference. It considers Gaussian Process models from the viewpoint of Machine Learning, a subset of artificial intelligence.

Long Term Forecasting